《A randomized polynomial-time simplex algorithm for linear programming》.pdf

A Randomized Polynomial-Time Simplex Algorithm for Linear Programming ∗ † Jonathan A. Kelner Daniel A. Spielman Computer Science and Artificial Intelligence Department of Computer Science Laboratory Yale University Massachusetts Institute of Technology ABSTRACT a theoretical standpoint, the study of linear programming We present the first randomized polynomial-time simplex has motivated major advances in the study of polytopes, algorithm for linear programming. Like the other known convex geometry, combinatorics, and complexity theory. polynomial-time algorithms for linear programming, its run- While the simplex method was the first practically useful ning time depends polynomially on the number of bits used approach to solving linear programs and is still one of the to represent its input. most popular, it was unknown whether any variant of the We begin by reducing the input linear program to a spe- simplex method could be shown to run in polynomial time cial form in which we merely need to certify boundedness. in the worst case. In fact, most common variants have been As boundedness does not depend upon the right-hand-side shown to have exponential worst-case complexity. In con- vector, we run the shadow-vertex simplex method with